Signal Blocks AI Assistants From Reading Your Private Messages
Signal, the encrypted messaging app, has barred AI agents including Google Gemini and Apple Siri from accessing messages unless the user has explicitly granted permission. Meredith Whittaker said the change is needed because Google and Apple are building agentic features deep into their operating systems. According to CyberScoop, Signal shared the change on Friday, and the app's leadership said it will block Apple and Google operating systems from taking screenshots of user messages that feed their AI features.
According to Whittaker, these agents work by scanning what is on your screen. In doing so, Signal data is captured and made available to the agentic systems so they can answer questions and act on a user's behalf. The risk is that private messages may end up on someone else's server, where they are no longer private. CyberScoop says she called this an existential threat to Signal and its mission, and said the environment is becoming increasingly hostile to privacy. CyberScoop adds that the announcement follows reports that Meta's AI agent Muse was able to access and read the text of Signal messages.
The updated Signal apps now block Siri on iOS, and Gemini and other agents on Android, from taking screenshots that the agents process with optical character recognition. CyberScoop names the targeted features as Siri's Onscreen Awareness and Gemini Intelligence, and says the restriction covers agentic systems built into iOS 27, and Android versions where Signal can detect an AI assistant trying to read its screen. Users who want to screenshot their own chats can turn the protection off in Signal's settings, but doing so may expose their messages, and those of their contacts, to AI agents. Signal builds on the screenshot blocking it introduced last year to defend against Microsoft Recall, which CyberScoop says could capture and store messages on Windows 11.
Whittaker admits this is only a partial fix, since agents may have other ways to reach app data through the device file system. She argues that Apple, Google and Microsoft give developers no easy way to restrict AI access without drastic compromises to the user experience, which is also why the protections are not yet available on macOS. According to CyberScoop, she criticised the vendors for integrating agents recklessly and called on them to provide simple technical tools and accurate documentation, including a clear way for developers to opt out of AI data access and processing. Apple, Google and Microsoft did not immediately respond to CyberScoop's requests for comment. Apple is adding finer controls to Device Control and Data Access and Full Disk Access in macOS 27.3, currently in public beta, which will mean more permission prompts. The Register notes this appears aimed mainly at third-party access rather than Siri itself.
Why it matters
Encryption in transit does not help if an operating system agent reads the decrypted screen, so the assumption that a secure messaging app keeps content private on the endpoint no longer holds. Signal admits its block is incomplete and says developers lack proper opt-out tools, so app-level fixes cannot be relied on alone. Leaders need to decide whether to control OS-level AI features on devices themselves.
What to do
- Review which phones in your organisation have Siri or Gemini screen-reading features enabled.
- Update Signal to the latest version on iOS and Android so the screenshot protection applies.
- Advise staff to keep Signal's screenshot protection turned on when discussing sensitive business matters.
- Assess whether device management policy can restrict OS-level AI agents on devices handling sensitive chats.